Integrative machine learning identifies ECM1 as a candidate autophagy-related biomarker for immune infiltration and prognosis in hepatocellular carcinoma

Abstract

Hepatocellular carcinoma (HCC) remains one of the leading causes of cancer-related mortality worldwide and exhibits substantial molecular heterogeneity, highlighting the need for reliable prognostic biomarkers and therapeutic targets. Increasing evidence suggests that autophagy plays a critical role in HCC progression. This study aimed to identify autophagy-related candidate biomarkers associated with prognosis and immune infiltration in HCC using an integrative machine learning and bioinformatics approach. Gene expression profiles from the GEO datasets GSE121248 and GSE64041 were analyzed to identify differentially expressed genes (DEGs) between HCC and normal liver tissues. Autophagy-related DEGs (ARDEGs) were identified by intersecting DEGs with the Human Autophagy Database, Molecular Signatures Database and GeneCards. Functional enrichment analyses (GO and KEGG) were performed to characterize their biological functions. Clinical information, including overall survival (OS), was obtained from the TCGA-LIHC cohort. Three machine learning algorithms, LASSO Cox regression, Random Survival Forest (RSF), and Support Vector Machine Recursive Feature Elimination (SVM-RFE), were applied to prioritize prognostically relevant genes followed by external validation. Integrated analyses of gene expression, clinicopathological characteristics, immune cell infiltration, transcription factor–miRNA regulatory networks, and drug–gene interactions were subsequently performed. Sixteen autophagy-related DEGs were identified and were significantly enriched in longevity regulation, FoxO signaling, and AMPK signaling pathways. Integrating the results from LASSO Cox regression, RSF, and SVM-RFE identified six candidate prognostic genes: ECM1, EZH2, KLF4, CDH1, FAS, and GABARAPL1. Among these, ECM1 emerged as the primary candidate for further investigation because of its distinctive association with autophagy-related gene signatures, epithelial–mesenchymal transition (EMT), immune infiltration, and overall survival. Notably, ECM1 exhibited reduced expression in HCC while displaying altered correlations with autophagy-related genes compared with normal liver tissue, suggesting a potential disruption of its regulatory network during hepatocarcinogenesis. This integrative computational study identified six autophagy-related candidate prognostic genes in HCC and prioritized ECM1 as a promising candidate for future mechanistic investigation. The observed associations between ECM1, autophagy, immune infiltration, and patient prognosis provide a rationale for further experimental studies to determine its biological role and evaluate its potential clinical relevance. The findings should be considered hypothesis-generating until validated in independent experimental and clinical studies.